mrkeyoor.com_
Tue 01 Sept 16:13 UTC
MemoryMCP servermeasured 19 Aug 2026

Memory

Memory is Anthropic's reference MCP server for persistent recall. It keeps a knowledge graph of entities, relations and observations in a local JSON file, and exposes nine tools an assistant can call to write facts down and read them back in a later session. It ships as @modelcontextprotocol/server-memory and runs over stdio.

BootBOOTS ✓We started it and initialize came back in 5.1s.
Tools9tools/list payload 10,750 bytes
Context tax~2,688estimated tokens for the whole tools/list, chars/4 estimate
Resources / prompts1 / counts returned by resources/list and prompts/list
Measured19 Aug 2026one run of the harness, clean environment, no credentials · how we test

How we started it

npx -y @modelcontextprotocol/server-memory
Verdict

Use it, but understand you are buying a note-taking habit, not a memory. We started it and it came up clean with no configuration at all, which is rare in this batch. The cost against you is that nothing writes to the graph unless your system prompt nags the model into calling create_entities and add_observations, and the retrieval tools do substring matching over a JSON file, so recall gets vague once the graph is large. For a single developer's assistant it is the cheapest persistence you can bolt on. For a team knowledge base, it is the wrong shape.

Use it if

You want your assistant to remember your stack, your preferences and your colleagues' names across sessions, and you're willing to write the system prompt that tells it when to save
You want a memory layer with zero credentials, zero network calls and a storage file you can open in a text editor and fix by hand
You're prototyping agent memory and want the reference implementation of the entity/relation/observation model before you build your own
You value being able to delete a memory permanently, which here means editing one file

Skip it if

You expect memory to happen automatically; without prompt instructions this server sits idle and you'll conclude, wrongly, that it's broken
You need memory shared across a team or across machines, because the store is a local file with no sync, no auth and no multi-writer story
Your recall needs are semantic; search_nodes matches text, so asking about 'auth' won't surface a node that only says 'login'
You're already running a hosted memory product, in which case you'd be paying the token cost of nine more tools for a second, worse copy
Your context budget is tight and you have no intention of storing anything, since the tool list is charged on every turn regardless

The bill for your context window

ToolEst. tokensDescribed?Required params
search_nodes~368yes (56 chars)1
open_nodes~363yes (57 chars)1
create_relations~338yes (106 chars)1
create_entities~335yes (51 chars)1
read_graph~326yes (31 chars)0
add_observations~279yes (64 chars)1
delete_relations~254yes (50 chars)1
delete_observations~237yes (65 chars)1

Nine tools cost about 2,688 tokens of tools/list, from a payload of 10,750 bytes, estimated at chars/4 and measured on our run. That is above the median for this batch and it buys you a fairly small vocabulary: create, delete and read, across entities, relations and observations. The weight is not in the descriptions, which are terse, but in the JSON schemas; create_entities and create_relations carry nested array schemas that dwarf their one-line descriptions. Compared with Notion, the heaviest server we measured in this batch, Memory is cheap. Compared with what it does, it is a real charge on every single turn, paid whether or not the model ever writes a note.

Setup reality

There is nothing to configure. We ran it with npx and a clean environment, no API keys, no config file, and the harness picked up no environment variable hints at all. It announced itself on stderr as the Knowledge Graph MCP Server and spoke protocol version 2025-06-18 as memory-server 0.6.3. The only knob worth knowing is the memory file path, which its docs describe as settable through an environment variable; leave it unset and you get a JSON file next to the installed package, which is a bad place for anything you want to survive an npx cache clear. Point it somewhere you back up. Boot was 5.1 seconds on our box, most of that npx fetching the package.

Questions people ask

Does the Memory MCP server remember things automatically?

No. It only stores what the model explicitly saves by calling a tool. You have to instruct the assistant in its system prompt to record facts and to read the graph at the start of a conversation, otherwise the graph stays empty.

Where does the Memory MCP server store data?

In a single local JSON file on the same machine, written over stdio by the server process. Its docs let you override the path with an environment variable, which you should do, because the default sits inside the installed package directory.

Is it worth 9 tools of context?

If you actually write memories, yes; about 2,688 tokens per turn is less than re-explaining your project every session. If you install it and never prompt the model to use it, you're paying that on every turn for nothing.

Can two people share one memory graph?

Not safely. There is no auth, no locking and no sync; it is one process reading and writing one file. Two agents pointed at the same file will overwrite each other's observations.

How this page is made: the server is spawned as an unprivileged user with a clean environment and no credentials, then asked for its tools, resources and prompts over stdio. Token figures are estimates at four characters per token, not a tokenizer count. One run, one machine. Corrections: contact the desk.

← All measured MCP servers